Sunday, 7 September 2025

Linn's SPACE Optimisation DSP feature for their DS/DSM digital audio streamers. Does it even work?

Linn Klimax DS/2 streamer getting tested.

A couple months back while evaluating the Linn Klimax DS/DSM streamers, I mentioned that the products have the capability to implement what Linn calls "SPACE Optimisation" (Speaker Placement And Custom Environment, we'll call it SO for short); available since 2015. Because the products are network-connected, they're always online and through the Linn Account webpage, you can check the settings for your streamer and activate the SO function.

I remember being intrigued by this function and the claims of significant room-correction capability. For this post, let's take some time to explore what Linn SO features. To start, here is what their literature says about the intent of the system:

Space Optimisation is a toolset available to all Linn DS/DSM system owners which uses highly sophisticated acoustic modelling to build up a complete picture of how your speakers, their placement, and the unique characteristics of your room interact to affect the sound you hear. Space Optimisation takes account of the characteristics of your speakers, such as the position of each drive unit and the response of bass reflex ports, and of the characteristics of your room, such as room dimensions, construction materials, and features like doors and windows, to accurately model these interactions. It then precisely identifies frequencies that are artificially distorted by your environment, and reduces their energy and decay time, to reveal the music that would otherwise be hidden. It also applies a delay to each speaker to ensure the content from each speaker reaches your ears at the same time.

As you can imagine from this description, SO is basically a DSP system that uploads customized filter settings to your Linn DS/DSM streamer. The key characteristic is that this system is not based on direct acoustic measurements of your system/room.

Rather, it takes information that you feed it (mainly room dimensions, and wall/floor/ceiling materials for estimates of sonic reflectivity), premeasured speaker info in Linn's database (see list here), and then calculates custom filters based on their mathematical model.

Currently SO is in its 3rd major revision. I believe earlier versions SO1 and SO2 used the Konfig app on your computer which has since been denoted as "legacy" software.

The user activates this feature by logging into the Linn Account page (aka "Kloud Konfig"), and constructs a replica of the room dimensions using "Room Designs".


Within that "Room Designs" tab, we can edit and review the various spaces that have already been added to the system like this:


As you can see, I've created some designs to test out already. The top 3 are just 5m-cubed rooms which are clearly suboptimal (and unrealistic), but because they are cubic, will accentuate room modes so we can examine the filters SO creates for them more clearly for comparison. Notice that the only difference between the 3 is the material of the walls/ceiling/floor - all concrete, dampened/suspended walls, or glass.

The bottom room design (highlighted yellow) is a model of my actual sound room with dimensions measured using a measuring tape, along with some extra detail with a built-in cabinet where I store CDs and other equipment. This is my attempt at seeing how SO performs in an actual room (details of this room with set-up suggestions summarized in a previous post).

As a quick recap then, recall that the listening room is obviously one of the most important "components" in our sound system and strongly determines what we hear. The reason is that sound waves interact with the boundaries. As discussed on this blog over the years and elsewhere, the Schroeder Frequency, defined as:

Schroeder Frequency: Nice graphic from SynAudCon. Many will use T60 (reverb time for sound to decay by 60dB) rather than T30 in this formula for Fs. Typically, the transition happens around 100-200Hz. Smaller rooms push the transition zone higher in the frequency range.

allows us to imagine the audible frequency spectrum separated into various zones. Room modes at lower frequencies dominate in the typical small sound room up to around 100Hz, beyond which gives way to a transition zone and then above that, dense modal overlaps result in a more continuous diffuse sound field rather than individual "standing waves".

I. The nature of Space Optimisation filters

Since Linn implements the feature based not on acoustic measurements but computed models mostly from room dimensions in their algorithm, logically, the most likely intent must be to try to calculate the low-frequency modes to create filters that smooth out the bass frequencies.

We can quickly test this idea with a simple 5m-cube room designed like this:

Note inset diagram of symmetrical positioning of speakers and listener in that "room" which I will use in many of the tests.

In the left panel we can see that the web-based software allows us to define ceiling features (flat or variable height), as well as the materials of the walls/ceiling/floor, and there's a facility to add features to the walls such as recessed portions or maybe a storage unit jutting out. That's pretty cool and will certainly add complexity to the calculations.

Once we map out the room, to create the SO filters, we select one of our Linn DS/DSM streamers to apply the room correction to and position the speakers and listener within that space. Depending on Linn's server load, from what I've seen, it'll typically take less than 5 minutes to calculate a new filter and download to the streamer.

Once the filters are calculated, they're saved and can quicky be recalled later:

Notice Space Optimisation algorithm 3.0.1 (Nov 29, 2019) being used.

Starting with that very simple 5m-cube "room", let's have the software calculate the filter for us, targeting generic speakers (there's a step where we define the output of the Klimax DS/2 as stereo front left/right channels which I won't show here).

Here's the filter it created for the left channel:


A few items of note:
1. The filter consists of parametric EQs.

2. They can only address room modes below 100Hz (limit for PEQ frequency looks like less than 90Hz). Regardless of room size, it's all done below 100Hz.

3. The EQ can only attenuate which is fine although we've seen over the years with other room correction DSP (like Acourate), we can also apply a small amount of "correction gain" like say up to +3dB if needed.
The room modes being addressed by the software in this simple cubic design follows the expected calculations for axial, tangential, and oblique modes such as from REW's Room Simulation module:


We can also use online tools like Bob Gold's Room Modes Calculator for this 5m x 5m x 5m room:


Correlating the modes above with the room simulation image from REW, we see that up to 100Hz in this 5m-cubic room, there are standing waves at 34, 68, 77, and 97Hz. Linn's EQs address the first three of these. I suspect the 97Hz mode was a bit too close to 100Hz. Examining other room designs, it looks like SO will only address modes with parametric EQ frequency up to 90Hz or so.

Space Optimisation allows us to identify the materials used for the boundaries of the room; this is their way to estimate reflectivity of walls. Here are overlaid frequency sweeps from the Linn Klimax DS/2 captured in REW with variant SO filters applied, changing only the wall material:


As you can see, the highest attenuation is being applied to the concrete room which correlates to that material being most reflective. Notice that the frequency response stretches out to over 35kHz indicating that the DSP is being performed internally at 96kHz in the Klimax DS/2, consistent with the 24/96 sweep.

Let's check at 192kHz samplerate with a white noise signal to make sure there's no downsampling being done:


Looks good with extension as expected for the 192kHz samplerate. Space Optimisation is able to function up to the maximum samplerate without downsampling; presumably the calculations are happening within their FPGA.

Let's now try a room design that's a bit more challenging and asymmetrical. Still in that 5m-cubed space but let's make the speakers and listener be positioned in the corners like this:


Let's see what the left and right channel filters look like now:


As you can see, since the speakers are placed at different distances from the walls, the algorithm adjusts the filters independently for each speaker. However, there's an obvious deficit - the amplitudes are not being balanced out even though clearly the speakers are at very different distances from the listener. A more complete room correction should have adjusted that left speaker to be a bit louder than the right in order to sound more balanced to the listener given the extra distance.

Furthermore, in the time domain, since the left speaker is many meters/feet further away from the listener compared to the right, there should be some relative delay added to the right channel to make sure the wavefronts arrive at the same time. Can we see this delay on an impulse response?


Surprisingly, no evidence of correction for time arrival done at all between the two channels! (Notice also the impulse response shows linear phase behavior.)

Here are the details of the filter parameters calculated by Linn SO for each channel:


Reviewing the frequencies represented, the algorithm accounts for all 3 types of modes - axial, tangential, and oblique.

So while we can appreciate that Linn Space Optimisation is adjusting to compensate room modes up to 90Hz or so, this is not like most calibration techniques as you might find even in an inexpensive AV receiver where you can tune speaker distance/delay and relative channel levels even when clearly the software is aware of the relative positioning of the speakers to the listener.

II. Actual room modeling and comparison with acoustically measured Dirac Live

Alright then, instead of a hypothetical 5m-cubed room, let's switch over to a real room:

Model this? To what level of detail and accuracy can we?!

I took out my tape measure for some actual numbers - here's my sound room mapped out for Linn SO including front L+R speaker height as recommended by the instructions and also listener height sitting at my sweet spot. I also made sure to include the right-side storage cabinet built into the room which is the largest fixed feature that would affect the sound. Materials were identified for the dry-wall side walls, dry-wall ceiling, wooden floor, and wood/glass depending on which side of the storage cabinet:


Here are the left and right filters Space Optimisation created for my sound room:


Relatively simple parametric EQs! Note the extra bit of complexity with the right filter likely accounting for the nearby wood/glass cabinet.

So does it work? Let's compare Linn Space Optimisation with actual measured Dirac Live (with Bass Control) which is what I normally listen with when using the Integra DRX-8.4 receiver. Here are overlaid left channel frequency sweeps with microphone placed at the main listening position:


I guess the Space Optimisation DSP marginally improved frequency response by dipping that 80Hz peak and slightly reduced the peak around 30Hz. For my room using these dimension, it's clearly far from impressive and definitely not precise enough even though I did spend a bit of time with the tape measure and the room design software.

Comparatively, you can see the improvement when I turn off SO and captured the sweep through the Klimax DS/2 using Dirac Live with Bass Control correction (dotted green), providing full-range equalization (using Dolby Atmos Music Target Curve down to 20Hz) with subwoofers turned on. Without the subs on, we can see the Dirac Live crossover roll-off for the individual left speaker (dotted red). Subwoofer calibration makes sense using a mic measurement-based approach. I don't know how Linn expects users to integrate subwoofers or if SO is helpful in any way.

III. "Practical" and "Ideal" positions and filter variations?

One interesting SO feature we can try is to identify an "Ideal" speaker position as well as define the actual "Practical" speaker location we will be using.

What does this do? Well, to have a look, let's go back to the hypothetical 5m-cubed room and see what changes whether we define an "Ideal" speaker position or not.
[Linn recommends using their vague "Tune Dem" procedure to find your ideal speaker location. In effect, what they call "ideal" appears to be a highly subjective choice.]


Here's what changed whether we set that ideal position or just used the "practical" position without an "ideal". Let's compare the right speaker filters (doesn't actually matter since speakers and listener placed symmetrically in the room):


As expected, since the "practical" speaker positions, listener position, and room dimensions are exactly the same whether we added an "ideal" reference or not, the filter's modal correction frequencies remain the same. The only differences are the subtle bandwidth and gain adjustments.

Let's see in a real-life situation what this does in my sound room. Suppose I have to move the speakers closer to the front wall by a few feet so the set-up is no longer an ideal, symmetrical layout:


First, let's see what the room sweep measurements looks like in the "Ideal" position with and without SO turned on:


Okay, so we see again that the effect of SO does provide some small benefits in pulling down a couple of the bass peaks like we saw earlier.

Now let's actually move my speakers to the "Practical" positions and repeat the sweeps with and without the "Ideal" reference to see how much difference this makes:


Hmmm, those are trivial differences below 100Hz.

For a better look at each channel, let's overlay "Practical only" vs. "Practical with Ideal reference" vs. actual "Ideal with SO turned on" curves:


Sure, we could say that the blue measurements which are the "Practical speaker position with Ideal reference" resulted in bass frequencies closer to the black "Ideal" position measurement. What this suggests is that by defining the "Ideal" position, the SO algorithm is probably calculating how a corrected frequency response at the "Ideal" location would look like and then sets the "Practical" location's EQ parameters to be closer to that rather than simply flat bass response.

Again, what I'm seeing here and in other related tests are just trivial differences and I would argue that the changes >100Hz are what truly define the listener's subjective preferences when using the "Tune Dem" approach. To be honest, I don't see how this "Practical" position + "Ideal" reference option is all that meaningful. It does do something though, so give it a try I guess and see if you like it through your Linn system. 🙂

IV. Frequency Response vs. Decay Time slider

Another setting is the balance slider between frequency response : decay time. The default is 80:20 in favor of frequency response. What if we turned that around to 20:80?


Here's what happened to the left speaker filter using that 5m-cubed "room" with concrete walls:


While I'm not sure what the principle behind this frequency response to decay time slider is since there are no actual acoustic measurements available for SO to determine decay time (ie. there's no reverb time or waterfall data for the algorithm to optimize), the results do indeed show what's described in the "Optimisation Preference" description.

With the default "flatter frequency response" bias (80:20), the software applies stronger low-frequency attenuation while the supposedly "shorter decay" filter suppresses the higher frequency modes. Since it's a frequency-dependent EQ change, I assume it'll be perceived like a mild tonal shift.

V. Environmental adjustment - Temperature & Humidity

Here's something cute they also added - temperature and humidity environmental settings! The default is a comfy 20°C and 50% humidity:


What do the filters look like if we cranked this up to 30°C/100% humidity and then down to 10°C/30%?


As expected, the differences are subtle; a few Hz change to the frequencies, and <1dB change with the gain values.

Notice that I only showed humidity down to 30% - there's a reason. 😉

Psssttt, Linn, I think you've got a bug in your system... The moment humidity drops below 27% at 10°C, we start seeing funky results like this:

Interesting peek at SO creating a complex 18-PEQ setting on the right.

Wow! That 1% drop in humidity from 27% to 26% at 10°C is a real doozy!? 🤣

Make sure you don't live in the very dry desert. Otherwise Space Optimiser will be very unhappy. 🥲 Anyone run into this issue before?! Clearly that filter on the right will kill your bass response.

VI. In summary...

I believe I've captured the core of what this system does. There are other things I didn't bother testing at length, such as some speakers apparently have been measured for SO (see list here); presumably this increases accuracy although when I fooled around trying different speaker options, I saw no difference in the PEQ filters created. I don't think we're told what features of a measured speaker are incorporated into the calculations - presumably info about the location of drivers and ports. I see this message (and this) suggesting that companies could independently measure and submit info to Linn to add to that list of speakers. I wonder if these are physical measures or acoustic data.

So, does it work? At best, I think it's a definite maybe. 😉

One could ask, given Linn's advertised claims, is Space Optimisation equivalent or even better than room correction based on filters created with data from a measurement mic in the room? I think the answer is clearly no. Certainly not better than the Dirac Live correction I compared SO to in my sound room.

Seriously though, how can Space Optimisation be very effective even if we meticulously measured the room dimensions down to sub-centimeter accuracy or tried our best to match our wall materials to the options available? An idealized, computer-modeled, room is simply never going to be exactly like the real thing if we're aiming for high-fidelity accuracy.

While I appreciate Linn SO's more sophisticated room modeling options than just rectangular or cubic designs, including designing stairways and angled ceilings, there will inevitably be all kinds of nuances in real life unaccounted for in the idealized room model. In the real world, actual domestic sound rooms are not devoid of furniture, carpets, doorways, small windows, decorations, even diffusion/absorption treatments (like bass traps) and other effects that would be impossible to account for accurately no matter how hard we tried!

In addition to the humidity bug, I'm disappointed that SO isn't able to account for what should be basic speaker output level adjustments nor does it add delay to account for relative distance from the listener even though the software has all the data to do so (even down to temperature and humidity for speed of sound calculations 😁)! This is despite Linn claiming "It also applies a delay to each speaker to ensure the content from each speaker reaches your ears at the same time." in the revision 3.7 manual from October 2023. Where is it that they're applying delay between the channels? Unless I've missed something, they might want to revise that claim at least based on my experience with the Klimax DS/2.

When it comes to fine-tuning the filter, such as adjusting this material absorption page:


Is the human ear able to do such a job anywhere near accurately?! What are those percentages of anyways? I would argue it's actually impossible. Even beyond the question of hearing acuity, this page is inherently limited because the "wood" material in my room would have a different characteristic whether it's flooring or the wood cabinet material; there's no allowance here for that kind of granularity.

Despite the decade that Linn SO has been available (since 2015), searching the Internet, I'm surprised I don't see any posted room measurements done by Linn enthusiasts or from the company showing the improvements one should expect. There's this interesting article about someone going around the world setting SO1 for others but nothing technical is discussed. Let me know if there have been other measurements posted that I missed.

Anyhow, if you have a Linn DS/DSM streamer, by all means, give SO a shot, work with your local dealer, try their "Tune Dem" procedure and see whether you like it. I can certainly appreciate the time and effort they must have expended making the software to create the on-line room-design interface and some of the little extra UI details. The idea of a simple non-technical room correction without setting up measurement microphones or looking at frequency sweeps is seductive. However, for those who have experienced more sophisticated DSP room correction (eg. FIR convolution systems) or even just used parametric EQ to compensate for measured room modes (eg. REW sweeps to identify problem frequencies then use software like Roon or external processor like Behringer DEQ2496), don't be surprised if the result with Space Optimisation isn't particularly impressive.

IMO, if you seriously want good room-correction, pull out your measurement mic (like the miniDSP UMIK-1) and get the job verifiably done. I would strongly disagree with Linn's claim that if we want good room correction, we should "drop the mic".
[Also see Linn's answer to "Why don't you use a Microphone in Space Optimisation?" for what sounds like idealistic rationalizations. It would be much more convincing if they just showed us how capable their room correction is in a real-life space with measured room responses than mere words! This would also give us an idea of how much room design detail is needed to achieve good results.

Given the comprehensive cloud-based foundation and DSP potential, something Linn should consider is adding an advanced option for users to plug in settings or fine-tune SO's calculated values. Users can edit the PEQ parameters themselves, maybe extend beyond the 100Hz limit as needed. The current "Custom Filters" settings with +/-3dB bass and +/-2dB treble shelf and subsonic filter options seem a bit weak. While at it, also add channel level controls (channel balance) and allow for relative channel delays.

At the price point for these DS/DSM streamers, providing options for advanced DSP customization might also empower Linn dealers to offer a much more comprehensive room optimization service for customers if the owner doesn't feel comfortable with the technical details. That's the kind of valuable service I would hope knowledgeable dealerships can offer for these "high-end" audio products beyond just selling stuff.]

--------------------

Latest Linn Klimax DS/2 firmware (July 30, 2025)...

I see that Linn had updated their firmware since my previous measurements so I proceeded with getting the Klimax DS/2 to the latest version (Davaar 112 - released July 2025) before running the SO tests above. I was wondering if this update improved the low-level "hash" I saw previously when playing a complex signal.


Space Optimisation turned off, of course. Typical 1/10-Decade Multitone 32 signal.

Nope, no change.
We still see the low-level noise in the playback unfortunately.

At this point I'm assuming that affected Linn DS/DSM models probably cannot be improved by firmware change since the company should already be well aware of the issue. My friend who owns this Linn DS/2 had reported the findings to the company more than a year ago.

With that, I think I understand as much as I need to about the Linn DS devices unless anyone wants to lend me their latest-and-greatest Linn Klimax DSM/3 Organik (MSRP US$39k) for testing. 😉

--------------------

I hope you're enjoying the music, dear audiophiles and had a great summer.

The Butchart Gardens, Victoria, Canada - late summer 2025.

Some new Christian McBride Big Band + Cécile McLorin Salvant off Without Further Ado, Vol. 1 (2025) - here's Cole Porter's "All Through The Night":


And for the rock/pop lovers - Rob Thomas' "Machine" from All Day Nights (2025); as usual, check out the multichannel/Atmos version if you can:

No comments:

Post a Comment